Sodium Valproate: Independent Evidence on Epilepsy & Pregnancy

Table of contents

• Evidence summary
• What sodium valproate is
• How it works
• What it is prescribed for
• What works and what does not
• Benefits by claim
• Risks and all side effects
• All interactions
• Who should avoid sodium valproate
• Dosage and how to take it
• Follow the money: who makes it and who funded the evidence
• Related research
• Frequently asked questions
• Sources and funding notes

Evidence summary

Independent evidence and credibility scorecard

Independence tiers: 1 = regulator, guideline body or government-commissioned review; 2 = academic or Cochrane team with no declared commercial funding for the work; 3 = independent authors with notable industry contributions or multi-company funding; 4 = manufacturer-funded or manufacturer-authored. Credibility runs from A (highest) to D.

Sources cluster in the UK (the regulator, review, guideline body and the trials are British), so the verdict reflects British regulatory and clinical practice. The strongest signal is convergence between opposing interests. The valproate maker's own product information, the regulator, a publicly funded Cochrane review (about 10% malformations) and a registry paid for partly by rival companies (10.3%) all agree on the size of the pregnancy harm. The favourable evidence for epilepsy comes from publicly funded trials, not from the maker.

What sodium valproate is

Valproic acid is the active molecule. In the UK it is used as the sodium salt and as a mixture of valproic acid and sodium valproate, sold as divalproex sodium or valproate semisodium. A magnesium salt is used in some other countries (Cumberlege report, para 4.1 footnote).

• Brand names: the MHRA lists Epilim and Depakote as the most commonly dispensed UK brands, with others including Convulex, Episenta, Epival, Kentlim, Orlept, Syonell and Valpal (MHRA, 2018). The French brand is Depakine (Sanofi 20-F). In the US, Depakote is marketed by AbbVie (US label).
• Legal status: prescription-only in the UK (SmPC). It was first licensed in the UK for epilepsy in 1972 (Cumberlege report, para 4.1).
• UK licence holder: Aventis Pharma Limited, trading as Sanofi, for both Epilim Chrono and Depakote (Epilim Chrono SmPC; Depakote SmPC).
• How many people: the Cumberlege review estimated about 27,000 women of childbearing age in the UK took valproate in 2020 (para 4.1).

How it works

The UK product information says the most likely mode of action is to strengthen the calming effect of the brain chemical GABA, by affecting how it is made or broken down (SmPC 5.1). That is a proposed mechanism, not a proven one. The same document shows valproate crosses the placenta in animals and humans and causes malformations in mice, rats and rabbits (SmPC 4.6, 5.3). What is proven in humans is the effect on seizures and mania in trials, and the harm in pregnancy in cohort and registry studies. We did not find an authoritative source that settles how it harms the developing baby, so we do not describe one.

What it is prescribed for

• Epilepsy (UK): generalised, partial or other epilepsy. For women and girls under 55, and for men under 55 starting treatment, only when there is no other effective or tolerated treatment (Epilim Chrono SmPC 4.1).
• Manic episodes in bipolar disorder (UK): Depakote is licensed for this, again only when no other effective or tolerated treatment exists. Continuing after the episode may be considered in people who responded (Depakote SmPC 4.1). NICE notes that other valproate products used for mania are off-label, although common in practice (NICE CG185).
• Migraine prevention (US): the US label also lists migraine prophylaxis and forbids that use in pregnancy (US label). The Cumberlege review notes it is also used for migraine and pain in the UK (para 4.1). We did not review migraine trials.

What works and what does not

Benefits by claim

Does it control generalised epilepsy?

Yes, and better than the main alternatives. In SANAD, valproate was better than lamotrigine for time to 12-month remission (hazard ratio 0.76, 95% CI 0.62 to 0.94) and, in idiopathic generalised epilepsy, better than both lamotrigine and topiramate for treatment failure. The authors still said the pregnancy risk must be weighed for women of childbearing age (Marson 2007). SANAD II randomised 520 people to levetiracetam or valproate. Levetiracetam did not meet the non-inferiority margin, the per-protocol analysis favoured valproate, adverse reactions were reported by 37% on valproate and 42% on levetiracetam, and levetiracetam was "dominated" by valproate in the cost-utility analysis, meaning it was no better and no cheaper overall. The authors said the results "inform discussions" about avoiding valproate in girls and women (Marson 2021). NICE reviewed the evidence and found that for time to treatment failure no drug performed better than valproate for generalised tonic-clonic seizures. Because of the MHRA rules it now offers lamotrigine, levetiracetam or valproate as a choice (NICE NG217).

Does it work for absence and focal seizures?

It is an option, not the first choice. NICE recommends ethosuximide first for absence seizures, then lamotrigine, levetiracetam or valproate. For focal seizures NICE recommends lamotrigine or levetiracetam first and keeps valproate as a third-line add-on, noting that its evidence there was less clear (NICE NG217).

Does it treat mania?

Yes, to a modest degree. The Cochrane review (25 trials, 3,252 people) found 45% responded on valproate versus 29% on placebo, and 56% versus 62% for valproate versus lithium (odds ratio 0.80, 95% CI 0.48 to 1.35). In children and adolescents the evidence was uncertain, and risperidone may have worked better (23% versus 66%, low-quality evidence) (Jochim 2019). NICE advises an antipsychotic first. It suggests adding lithium if that fails, and adding valproate if lithium is ineffective or unsuitable (NICE CG185).

Does it prevent bipolar relapse?

The evidence is thinner than many people assume. Cochrane found six trials with 876 people, and valproate beat placebo for staying in the trial without a mood episode (relative risk 0.68; number needed to treat 8) with no difference from lithium (Cipriani 2013). BALANCE found that relapse needing intervention occurred in 69% on valproate alone, 59% on lithium alone and 54% on both. Lithium beat valproate (HR 0.71, 0.51 to 1.00), and the combination beat valproate (HR 0.59, 0.42 to 0.83) (BALANCE 2010). See our lithium review.

Does it help bipolar depression?

Not convincingly. The only pooled support is a 2024 overview of systematic reviews funded by Abbott Laboratories, which reported benefit over placebo from 7 small trials (Mari 2024). NICE recommends other treatments first (CG185). See bipolar depression.

Risks and all side effects

Side effects listed in the UK product information, by frequency (SmPC 4.8):

• Very common (more than 1 in 10): nausea; tremor. The SmPC also says valproate very commonly causes weight gain, which can be marked and progressive (4.4).
• Common: liver injury; vomiting, diarrhoea, stomach pain, gum overgrowth, mouth ulcers; sleepiness, stupor, memory impairment, headache, involuntary eye movements, movement disorders, convulsions; confusion, hallucinations, aggression, agitation; low blood sodium; low platelets and anaemia; hair loss (often temporary) and nail problems; bleeding; painful periods; urinary incontinence; deafness (cause not established).
• Uncommon: pancreatitis; encephalopathy, coma, reversible parkinsonism, unsteadiness; raised male-type hormones (acne, excess hair, hair loss in a male pattern); SIADH (water retention); pancytopenia and leucopenia; angioedema and rash; absent periods; vasculitis; kidney failure; low body temperature; bone thinning, osteoporosis and fractures with long-term use; worsened seizures; eosinophilic pleural effusion.
• Rare: reversible dementia with brain shrinkage; high blood ammonia; obesity; underactive thyroid; bone marrow failure; Stevens-Johnson syndrome, toxic epidermal necrolysis and DRESS; polycystic ovaries; male infertility; lupus; muscle breakdown (rhabdomyolysis); kidney tubule damage (Fanconi syndrome); abnormal clotting tests; myelodysplastic syndrome.
• Behaviour in children: aggression, agitation, hyperactivity and learning problems are mainly seen in children.
• Suicidal thoughts and behaviour: reported with anti-seizure medicines in general, and a small increased risk across these drugs in a pooled analysis of trials; the data do not exclude it for valproate (SmPC 4.4).

NICE adds that long-term valproate can reduce bone density and raise the risk of osteomalacia, and that supplementation should be considered for people at risk (NICE NG217). For related liver and pancreas conditions, see our guides to drug-induced liver injury, acute liver failure and acute pancreatitis.

All interactions

The UK product information lists these interactions (SmPC 4.5). The US label lists similar ones (US label).

Drugs that change valproate levels

• Carbapenem antibiotics (imipenem, meropenem): valproate levels can fall 60 to 100% within two days, sometimes with convulsions. Combining them is not recommended.
• Enzyme-inducing anti-seizure drugs (phenytoin, phenobarbital, carbamazepine) and rifampicin: lower valproate levels. Felbamate, cimetidine and erythromycin: raise them.
• Oestrogen-containing contraceptives: may lower valproate levels and effect. Valproate does not weaken hormonal contraception.
• Others that lower levels: protease inhibitors (lopinavir, ritonavir), cholestyramine, metamizole, methotrexate, and the antimalarials mefloquine and chloroquine, which may also lower the seizure threshold.
• Aspirin and other salicylates: raise free valproate and the liver risk; avoid in under-16s and in under-3s especially.

Drugs whose effects valproate changes

• Lamotrigine: valproate slows its breakdown, raising toxicity and the risk of serious skin rash; lamotrigine doses need reducing.
• Phenobarbital, primidone, phenytoin, carbamazepine: raised effects or toxicity, including sedation.
• Warfarin and other coumarins: stronger blood-thinning effect; clotting tests need close monitoring.
• Topiramate and acetazolamide: linked to encephalopathy and high ammonia.
• Clozapine, olanzapine, quetiapine: more neutropenia (and myocarditis with clozapine); valproate may lower olanzapine levels.
• Antidepressants, MAOIs, benzodiazepines, other psychotropics: effects may be increased.
• Others: propofol, zidovudine, nimodipine, rufinamide, and pivalate-containing medicines (carnitine loss). Valproate has no effect on lithium levels.
• Cannabidiol: in trials, ALT above 3 times the upper limit occurred in 19% of patients taking both. See our CBD review.
• Alcohol: not recommended during treatment.

Who should avoid sodium valproate

The UK product information lists these as contraindications (SmPC 4.3):

• Pregnancy, unless two specialists independently document that no other effective or tolerated treatment exists.
• Women and girls able to have children and under 55, unless two specialists agree and the Pregnancy Prevention Programme conditions are met. These are a pregnancy test, highly effective contraception, a signed annual risk form and at least yearly specialist review (MHRA 2018).
• Active liver disease, or a personal or family history of severe, especially drug-related, liver dysfunction.
• Urea cycle disorders, porphyria, POLG mitochondrial disorders (including suspected cases in children under 2), and uncorrected primary carnitine deficiency.
• Hypersensitivity to valproate or any ingredient.

Extra caution applies to children under 3 (liver risk), people with lupus, kidney impairment, and people taking other anticonvulsants (SmPC 4.4). Breastfeeding: valproate passes into milk at 1 to 10% of maternal blood levels, and blood disorders have been seen in breastfed infants (SmPC 4.6). Men: it can impair male fertility, partly reversibly. The MHRA advises men and their partners to use effective contraception during treatment and for at least 3 months after stopping, and not to donate sperm in that period. Men planning a family within a year are advised to see a specialist about alternatives (MHRA, 2024). The US label recommends folic acid before conception and in pregnancy for women taking valproate, but says it is not known whether this lowers valproate's risk (US label). NICE advises that a woman or girl of childbearing potential already taking valproate for bipolar disorder be advised to stop it gradually under medical supervision (NICE CG185 1.10.27).

Dosage and how to take it

These are label and trial doses, not advice for any person.

• UK label, epilepsy (Epilim Chrono, adults): start at 600 mg a day, increasing by 200 mg every three days until controlled, usually 1,000 to 2,000 mg a day (20 to 30 mg/kg), with a maximum of 2,500 mg. Children over 20 kg: start at 400 mg a day, usually 20 to 30 mg/kg a day (SmPC 4.2).
• UK label, mania (Depakote): start at 750 mg a day in 2 to 3 divided doses; the usual daily dose is 1,000 to 2,000 mg, and doses above 45 mg/kg need careful monitoring (Depakote SmPC 4.2). The US label has a maximum of 60 mg/kg a day (US label).
• Dose studied in SANAD II: advised starting maintenance of 500 mg twice a day for those aged 12 and over (Marson 2021).
• Prolonged-release tablets: swallowed whole, not crushed or chewed. Taking tablets with or after food helps early stomach upset (SmPC 4.2, 4.8).
• Monitoring: blood levels are not routinely needed (reported range 40 to 100 mg/L); NICE says to measure them only if the drug seems ineffective, adherence is poor or toxicity is suspected. For bipolar disorder NICE advises weight or BMI, a full blood count and liver tests before starting (NICE CG185).
• Brand switching: the SmPC notes NICE has advised that switching between valproate preparations is not normally recommended because levels can vary (SmPC 4.4).
• Dispensing in the UK: since 2023 valproate is supplied in whole packs with warnings on the box (MHRA collection). It must be started by a specialist, and in UK practice two specialists for anyone under 55.

Follow the money: who makes it and who funded the evidence

Who makes and sells it

• Sanofi (France): the UK licence holder for Epilim and Depakote is Aventis Pharma Limited, trading as Sanofi, of Reading, Berkshire (emc). Sanofi's registered office is in Paris. L'Oréal is its largest shareholder, holding 7.2% of the share capital and 13.1% of voting rights after a February 2025 share buy-back (Sanofi 20-F 2024).
• History of the UK licence: the original licence was held by Pharmacy Products (UK) Ltd, a joint venture of the Labaz Group and Reckitt & Colman. Sanofi acquired Labaz in 1981, and the Cumberlege review calls Sanofi the principal manufacturer of sodium valproate (report, para 4.5 and footnote).
• AbbVie (US): markets Depakote in the US; the label gives its address as North Chicago, Illinois (US label).
• Generics: many other firms sell valproate products. We did not count them or trace their owners. We also did not find a citable figure for valproate sales.

Who funded the evidence

• Public money: SANAD and SANAD II (NIHR, which is funded by the Department of Health and Social Care; NIHR); the 2023 Cochrane malformation review (NIHR); NEAD (US National Institutes of Health and UK Epilepsy Research Foundation).
• Mixed: SANAD was publicly funded by NIHR but reports industry contributions that we could not quantify. EURAP is paid for by eight makers (including Sanofi-Aventis and several makers of rival drugs) plus non-profit bodies. BALANCE was paid for by a US research institute and Sanofi-Aventis.
• Maker-funded: the 2024 BMJ Open overview of valproate in bipolar disorder was funded by an Abbott Laboratories grant. The Cochrane 2013 maintenance review's lead authors reported Sanofi-Aventis honoraria and BALANCE roles.
• Regulators: the Cumberlege report said the MHRA's medicines work was funded primarily by pharmaceutical industry fees, while the MHRA says it keeps independence through conflict-of-interest policies (para 2.59). That was the position in 2020; we did not find a more recent figure.

Documented events

• EU review: the 2013 to 2014 review was started at the MHRA's request. A second review began in March 2017 at the French regulator's request and led to the 2018 programme (EMA 2014; EMA 2018).
• UK court case: claims by more than 100 children against Sanofi Synthelabo began in 2004 with legal aid. Legal aid was withdrawn in 2006, restored, and withdrawn again in October 2010 a few weeks before trial. The Cumberlege report records that the Legal Services Commission was advised the claim was unlikely to succeed (para 4.87). At the end of 2024, Sanofi reported one pre-action case in Great Britain and one civil claim in Northern Ireland.
• Cumberlege on Sanofi: the panel said it had seen no evidence of specific failures by Sanofi to act within the regulatory framework of the time, but that merely complying was not enough. It concluded Sanofi has an "ethical obligation" to contribute to a payment scheme. Sanofi was the only one of the three manufacturers reviewed to attend an oral hearing (paras 4.103 and 4.104).
• France, civil: at 31 December 2024, 79 families (133 people) had civil claims against Sanofi's French affiliate. A Paris court ruled on 5 January 2022 that a class action was admissible and retained Sanofi's liability for malformations between 1984 and January 2006 and for neurodevelopmental disorders from 2001 to January 2006. Sanofi appealed. Four first-instance cases had ruled for plaintiffs and two for Sanofi, all under appeal. Sanofi states it cannot assess the outcome (Sanofi 20-F 2024).
• France, criminal: an investigation began in May 2015. Sanofi's French affiliate was indicted in January 2020 for aggravated deception and involuntary injuries, and in July 2020 for involuntary manslaughter. The French medicines agency ANSM was indicted in November 2020. In 2022 and 2023 courts ruled certain complaints time-barred. Indictment is not a finding of guilt, and Sanofi's filing records no verdict (Sanofi 20-F 2024).
• France, state scheme: a public compensation scheme (ONIAM) has found Sanofi's affiliate liable in whole or part in several opinions. Sanofi disagrees, has not offered compensation to those paid by ONIAM, and is contesting ONIAM's reimbursement orders. In January 2025 the Paris Administrative Court of Appeal retained the State's liability in five cases and, for births from 2006, found no fault by Sanofi. For births in 1999 and 2002 it set Sanofi's share at 50% (Sanofi 20-F 2024).
• Elsewhere: Sanofi reports claims in Switzerland (11 families, 17 people), Spain (7 actions, 15 children), Belgium, Ireland, and Northern Ireland.
• UK government response: the government answered the Cumberlege report on 21 July 2021 (gov.uk). The Patient Safety Commissioner's report of 7 February 2024 recommended a two-stage redress scheme for valproate and pelvic mesh. Her website then said the government had made no final decision and no scheme was set up (Patient Safety Commissioner). We could not confirm any later decision.

Related research

For wider context, see these Pure City Research guides:

• Lithium: bipolar, suicide and kidney risk
• Bipolar depression: assessment, treatment evidence and safety
• Clonazepam: panic, epilepsy, side effects and withdrawal
• Lorazepam: anxiety, seizures and withdrawal
• CBD: anxiety, sleep, liver risk and law (interacts with valproate)
• Drug and supplement induced liver injury
• Acute liver failure
• Acute pancreatitis
• Self-harm: assessment, therapy and safety

Frequently asked questions

Sources and funding notes

What we could not verify: the full funding statements of the 2019 and 2013 Cochrane bipolar reviews (Europe PMC lists NIHR grants for both); the size of industry funding in SANAD (2007); whether Cochrane later ruled the 2014 neurodevelopment review non-compliant with its conflict policy; NICE and EMA funding statements (not on the pages we opened); any rulings or UK redress decisions after the sources listed; trial evidence for migraine; the Sanofi sales figure for valproate; and the exact mechanism of teratogenicity. We did not review news published after Sanofi's 2024 annual report.

1. MHRA Drug Safety Update, April 2018 (volume 11, issue 9) — regulator; UK government agency (see funding note under the MHRA entry below).
2. MHRA: Valproate use by women and girls (guidance, updated September 2025) — regulator; UK government agency.
3. MHRA: New valproate safety measures apply from 31 January (2024) — regulator; UK government agency.
4. MHRA Drug Safety Update, 5 September 2024: valproate use in men — regulator; the underlying study was requested by the European regulator and funded by a consortium of valproate makers (Colas 2025 funding statement).
5. MHRA: Valproate safety measures (collection) — regulator; timeline from 2021 to 2025.
6. MHRA organisation page — says it is an executive agency sponsored by the Department of Health and Social Care.
7. First Do No Harm: report of the Independent Medicines and Medical Devices Safety Review (2020) — government-commissioned independent review; chaired by Baroness Cumberlege. The same report states MHRA medicines funding came mainly from industry fees (2020).
8. Government response to the review (21 July 2021) — UK government.
9. Patient Safety Commissioner: redress page and news item on valproate use by men (6 September 2024) — the commissioner's office recommended by the Cumberlege review; funding statement not opened.
10. Epilim Chrono 500 mg SmPC (emc) — written by Sanofi; MHRA-approved.
11. Depakote 500 mg tablets SmPC (emc) — written by Sanofi; MHRA-approved.
12. Depakote (divalproex sodium) US prescribing information, AbbVie — written by AbbVie; FDA-approved.
13. NICE NG217: Epilepsies in children, young people and adults (updated August 2026) — UK public body; no funding statement opened.
14. NICE CG185: Bipolar disorder (updated September 2025) — UK public body; no funding statement opened.
15. EMA: valproate referral (2014 outcome) — EU regulator; funding statement not opened.
16. EMA PRAC recommendation, 9 February 2018 — EU regulator; review requested by France's ANSM.
17. Marson et al., SANAD, Lancet 2007 — NIHR HTA programme; industry contributions reported but not quantified here (full text).
18. Marson et al., SANAD II, Lancet 2021 — NIHR HTA programme; some authors report UCB and Sanofi grants or fees.
19. Bromley et al., Cochrane 2023: monotherapy treatment of epilepsy in pregnancy, malformations — grants from NIHR, Epilepsy Research Institute UK and the Novo Nordisk Foundation; minor author ties to UCB, GSK, Eisai and Sanofi-sponsored events.
20. Bromley et al., Cochrane 2014: neurodevelopmental outcomes in the child — NIHR grants listed in Europe PMC; authors reported work on Sanofi- and UCB-funded studies.
21. Tomson et al., EURAP, Lancet Neurology 2018 — Bial, Eisai, GlaxoSmithKline, Janssen-Cilag, Novartis, Pfizer, Sanofi-Aventis, UCB, Netherlands Epilepsy Foundation, Stockholm County Council.
22. Meador et al., NEAD study, Lancet Neurology 2013 — US National Institutes of Health and UK Epilepsy Research Foundation; some authors report industry grants and expert-witness work (full text).
23. Christensen et al., JAMA 2013: prenatal valproate and autism — Danish registry study; European Research Council and US CDC grants listed in Europe PMC; lead author reported fees from UCB Nordic and Eisai.
24. Jochim et al., Cochrane 2019: valproate for acute mania — Europe PMC lists an NIHR grant; authors report no conflicts.
25. Cipriani et al., Cochrane 2013: valproate in maintenance of bipolar disorder — NIHR grant listed in Europe PMC; authors report Sanofi-aventis honoraria and BALANCE roles.
26. BALANCE investigators, Lancet 2010 — Stanley Medical Research Institute and Sanofi-Aventis.
27. Mari et al., BMJ Open 2024: overview of valproate in bipolar disorder — grant from Abbott Laboratories.
28. Sanofi Form 20-F for 2024 (US SEC) — Sanofi's own filing; litigation, shareholder and address facts.
29. NIHR: about us — says it is funded by the Department of Health and Social Care.

Educational research reviewed 6 October 2026. This article is not medical advice; talk to a qualified clinician or pharmacist before starting, stopping or combining treatments.